They did not explicitly model the probability of a draw—even though it was the most common outcome among top competitors. "As a result, these systems may produce stagnant ratings," observed Glickman, ...